Thank you so much really appreciate the kind words...
You are right.Context loss during incident response is the problem we’re aiming to solve deeply, especially for teams that live in Slack and don’t want to chase dashboards.
Let me try to answer your great questions:
1. Conflicting advice or noisy threads:
Right now, RobinRelay summarizes past replies per alert occurrence (thread), then aggregates them across all past occurrences of the same (e.g., “CPU spike on api-server”).
If conflicting advice exists (e.g., restart vs scale up), we preserve each cause separately and display them like:
“In the past, this alert was caused by:
– autoscaler OOM restart (3x)
– DB connection saturation (2x)”
We’re working on tagging which ones were “confirmed fixes” vs just investigative noise and will open that to user feedback soon.
Alert variants:
We normalize alerts our algorithm similar to finger prints for vector db.
We’re exploring multiple ways like fuzzy hashing, LLM's, custom way to detect meaningful variants (e.g., same service, different root cause) and eventually flag them as “related but distinct”.
Trend detection (e.g. precursor alerts):
Yes! This is on our roadmap.
We want to show timelines like:
If you're seeing these pain points first-hand, I’d love to chat or hear how your team handles it today — this kind of feedback shapes the product every week.
Thanks again for the thoughtful comment